Abstract: Immersive digital twin technology for mechanics experiments which fusing structural entities in real world with mechanics models in virtual space, not only is conducive for the experimenter to comprehend the deformation of structures subjected to loads but also contributes to the balanced regulation, sharing, and conservation of computational resources. Aiming at the problems of automation, of accuracy, and of the stability of mechanics experiments, proposed is a distributed digital twin system architecture based on the behavior sensing in edge hardware, on the data storage in cloud server and finite element computation in fog server, and on the virtual model visualization in Unity/Web/ Mixed-Reality of HoloLens platform, respectively. Besides, the indexes, such as mapping rate, fluctuation rate, and behavior error, are proposed to evaluate the mapping accuracy between the physical structure and the virtual model in the time domain as well as in the space domain. Further, three immersive digital twin systems of a cantilever mechanics experiment based on these proposed architectures were built. Systematic analysis shows that all the three architectures can achieve automatic, accurate, and stable immersive mechanics experiments. Among them, Web platform achieves a slightly higher refresh rate and a smaller fluctuation rate and is easier to access on, while Unity platform is more convenient and realistic for the 3D digital scene reconstruction, and the mixed reality of HoloLens obtains a more immersive and interactive experience.